Metrics Deduplication

[Adso Castro]

Hey all,

I have a curious question here:

I have 3 Prometheus replicas running under my Prometheus-Operator (plus Thanos) inside a Kubernetes Cluster. When I query something within the range of 3 or 6+ hours, I get the same metric x 3.

Is that correct or should I get a single metric already refined by Thanos?

Thank you.

[Mat Arye]

I am not positive but I think this depends on how external_labels and Thanos are set up. Namely, in Thanos the "query.replica-label" determines how deduplication is done. See https://thanos.io/components/query.md/

Thank you.

[Adso Castro]

That sure helps a lot. Thank you.